Why DDR5 ECC Matters: Protecting Data from Silent Hardware Failures
ECC and DDR5
I explore how ECC memory prevents data corruption caused by hardware errors, noting that many Firefox crashes and filesystem issues stem from faulty RAM. While DDR5 introduces on-die ECC, I argue it doesn't replace traditional motherboard-level protection. I also highlight the confusing market for EC4 versus EC8 modules and urge for wider ECC adoption to reduce wasted developer time and data loss.
"We need ECC RAM to be more widely used, ideally with government action or taxes on non-ECC RAM to make it more expensive than ECC when bought new."
HN discussion
- DDR5 on-die ECC was implemented to prevent reliability degradation from increased error frequencies rather than to improve upon older generations, and it fails to protect against electrical noise during data transmission between the CPU and memory modules.
- Overclocked 'gaming' DDR5 DIMMs lack manufacturer guarantees for stability at advertised speeds, making them unsuitable for professional workloads where silent errors caused by environmental factors like heat can occur daily.
- While some argue on-die ECC might suffice for home servers, practitioners emphasize that full system-level ECC is critical for detecting physical issues like oxidized contacts or worn-out modules that on-die correction cannot address.
- The cost of ECC memory represents a negligible fraction of total system cost, making the trade-off of accepting data corruption to save a few dollars an unreasonable risk for any system handling important data.